PLG-SUNetT: Parallel Local-Global Swin-UNet Transformer for Brain Tumor Detection

The precise segmentation of brain tumors in magnetic resonance imaging (MRI) pictures is necessary for the efficient planning and monitoring of treatment during the process. In many computer vision applications, the Swin transformer performs better than the U-net architecture when it comes to brain tumor segmentation. The incorporation of the self-attention module into the Swin transformer provides global visual dependency. In this paper, a proposal is made for a Parallel Local-Global Swin-UNet Transformer (PLG-SUNetT), which is comprised of the hierarchical Swin Transformer. There is an increase in explainability in the decision-making process for the diagnosis of brain tumors using the Swin transformer-based U-shaped encoder-decoder with skip connection technology. The parallel process of local and global feature integration is the reason why the Modified Self-Attention Module (MSAM) is so efficient in terms of computing. The method that has been proposed is capable of, in addition to segmenting the tumor, classifying the type of tumor. Regarding the evaluation of the proposed model, the two datasets that are accessible publicly, namely Figshare and Kaggle multi-class datasets, were utilized. The PLG-SUNetT model that has been developed gets results that are superior to the methods that are currently considered to be state of the art.